摘要
目前,国内外对于变压器三相不平衡损耗的研究主要是使用传统公式法,精确度不高且需要较多的变压器内部参数。为提高变压器的三相不平衡损耗精确度,提出一种基于纵横交叉优化BP神经网络的损耗评估方法。该方法针对传统BP神经网络中的不足进行优化,将传统的随机初始权值阈值改为经过纵横交叉算法得到的最优权值与阈值,并将最优值代入训练模型中,在训练过程中传统BP神经网络易于陷入局部最优的不足也将因为纵向交叉的优化而得到解决,最终得到基于CSO-BP神经网络的变压器三相不平衡损耗评估模型。将该模型的损耗评估结果、公式法计算结果及实际实验得出的结果与现场实验数据进行对比,结果表明,基于CSO-BP神经网络的变压器谐波损耗评估模型得出的结果更接近实验值,具有良好的工程应用前景与价值。
At present,the traditional formula method is mainly used for the study on the three-phase unbalanced loss of transformers at home and abroad,and the method is low in accuracy and requires more internal parameters of transformers.In order to improve the accuracy of the three-phase unbalanced loss of transformers,a loss evaluation method based on the vertical and horizontal cross optimization BP neural network is proposed in this paper.This method is to optimize the shortcomings of the traditional BP neural network.The traditional random initial weight threshold is changed to the optimal weight and threshold obtained by the vertical and horizontal cross algorithm,and the optimal value is substituted into the training model.In the training process,the shortcoming of the traditional BP neural network that is easy to fall into the local optimum will also be overcome because of the optimization of the vertical cross.Finally,we get the three-phase unbalanced loss evaluation model of transformer based on CSO-BP neural network.Comparison of the loss evaluation results of the model,the calculation results of formula method with the results of the actual experiment suggests that the results of the harmonic loss evaluation model of transformers based on CSO BP neural network are closer to the experimental values,which has good engineering application prospect and value.
作者
陈子辉
吴智影
刘贺
孟炘
吴非
CHEN Zihui;WU Zhiying;LIU He;MENG Xin;WU Fei(Jiangmen Power Supply Company,Guangdong Electric Power Company,Jiangmen 529000,Guangdong,China;School of Automation,Guangdong University of Technology,Guangzhou 510006,Guangdong,China)
出处
《电网与清洁能源》
2020年第7期57-63,共7页
Power System and Clean Energy
基金
国家自然科学基金项目(61876040)。
